WhatsApp Bot for Business in Kazakhstan: Real Scenarios

Text: On July 1, 2025, Meta transitioned the WhatsApp Business Platform from charging for 24-hour dialogues to charging for each delivered template message. Previously, one payment covered any number of messages within a daily window; now, the company pays separately for each delivered template, and the price depends on the category and recipient's country: marketing messages cost approximately $0.025 to $0.137, while service and utility templates range from $0.004 to $0.046. However, responses within a 24-hour window opened by the client remain free. Another figure rarely mentioned in bot discussions: in 2022, over 80 percent of Kazakhstani people named WhatsApp as their primary messenger, but by 2025, Telegram confidently holds the top spot in the country. This immediately changes the conversation framework: WhatsApp is no longer the default sole channel and has become one of several, for which companies must pay consciously.

At West Star Ltd, we have been integrating messengers into business accounting systems for several years and see both extremes: companies either expect miracles from the bot and become disappointed or are afraid to touch it due to the unclear payment model. In this article, we will factually analyze what exactly is automated through WhatsApp, how the new pricing model is structured, where the boundary between free and paid lies, how to connect the bot with 1C, and in which situations this channel will not work. Without promises that the bot will replace the sales department.

WHAT IS A WHATSAPP BOT AND WHAT IT IS NOT

The first thing to understand is that the term "WhatsApp bot" hides two completely different things. There is the free WhatsApp Business app for smartphones — it suits microbusinesses and has limited capabilities: auto-reply during non-working hours, greetings, quick replies, and a catalog. This is not a bot in the engineering sense but a set of templates within a regular app, and serious logic cannot be embedded there.

True automation exists on another level — this is the WhatsApp Cloud API, a programmatic interface from Meta. It allows receiving incoming messages via webhook, processing them on your server, accessing databases, and responding automatically. Access is obtained either directly through the Meta platform or through partner providers. Only at this level can you build a bot that checks product stock, generates an invoice, or registers an application.

From this follows a practical conclusion. If someone offers you mass messaging through a regular number and app on a phone from a computer, it is almost always a gray scheme: mass automatic messages outside the official API violate the rules and lead to number blocking. Legal automation is always an official API and verified templates.

HOW PAYMENT IS STRUCTURED AFTER JULY 2025

The new model is simpler than it seems if you remember one concept — the client service window. When a person writes to your WhatsApp, a 24-hour window opens. Within it, you can respond with any free messages — text, images, files — and this is free. This is why support scenarios where the client initiates the dialogue cost almost nothing.

You have to pay for template messages. These are pre-approved texts with Meta that businesses send outside the window — for example, to remind about an appointment or inform about order status. Templates are divided into categories: marketing, service, and authentication. Each delivered template message is now charged separately, whereas before July 2025, one dialogue was paid for in full regardless of the number of messages.

For budget planning, this means simple logic. It is cheaper and more sustainable to build scenarios where the client writes first, and the bot responds within the free window. It is more expensive to mass initiate dialogues with marketing templates. Any template undergoes Meta moderation before use, and approval for verified businesses is usually instant, while in other cases, it takes up to a day. This should be factored into launch timelines.

Let's explain with a simple calculation why structure is more important than the rate. Suppose you have a thousand inquiries a month, and in eight out of ten cases, the client writes first. Then eight hundred dialogues close within the free window and cost almost nothing, and you will mainly pay for service notifications and rare marketing templates. If you change the logic and start most dialogues yourself, the same thousand inquiries turn into a thousand paid templates, and the bill increases several times with the same result. Therefore, the first question when designing a bot is not "which rate to choose," but "how to make the client write first": a form on the website with a button to start a conversation, a QR code at an offline point, a link in the email signature. Each such entry point opens a free window and reduces the channel's final cost.

SCENARIOS THAT ARE ACTUALLY AUTOMATED

In our practice, it is not the flashy but the boring repetitive tasks that pay off best. The first line of support: the bot answers typical questions about working hours, address, availability, and delivery conditions, while complex inquiries are passed to a live operator with already gathered context. This removes up to half of routine dialogues from people.

Order and booking statuses are the second reliable scenario. Order confirmation, shipment notification, and visit reminder a day before it belong to the service category and fit well within platform rules because they provide value, not advertising. Clients expect such messages and rarely complain about them.

Application acceptance and lead qualification work where the person initiates the dialogue: the bot asks several questions, records the answers, and passes a ready card to the manager. Payment acceptance is usually implemented in Kazakhstan through a link to a payment page directly in the conversation, not within the messenger itself. Finally, one-time codes and confirmations through the authentication category are a simple and predictable price scenario.

Special mention goes to reminders and funnel returns. Notifications about an abandoned cart, promotion expiration, or service renewal fall into the marketing category and require calculation, but with careful frequency, they pay off. The main rule here is not to turn a useful channel into spam, as unsubscribes and complaints hit the number's reputation harder than a one-time benefit from a mailing.

What not to expect from a bot: it poorly handles long unstructured negotiations, cannot make decisions on non-standard claims, and does not replace a person where voice, empathy, and bargaining are needed. A good bot handles a clear flow of questions and carefully passes the rest to people.

INTEGRATION WITH 1C

Messaging in a messenger brings little benefit if the data does not enter the accounting system. Value appears when the bot is connected to 1C. Architecturally, it looks like this: the WhatsApp Cloud API via webhook sends incoming messages to an intermediary server, which accesses 1C for data and returns the response back to the conversation.

The cleanest way to connect with 1C is through its web services. The platform can publish data via the OData protocol, allowing the bot to read stock, prices, and order statuses without manual uploads and without direct database access. The bot asks for product availability — the intermediary queries 1C, gets the current figure, and responds to the client in a second. An application from the conversation turns into a document in the system, and the status of this document is then returned to the client with an automatic notification.

In practice, it is convenient to build the connection so that the intermediary is responsible not only for reading but also for writing. When a client confirms an order in the conversation, the intermediary creates a customer order in 1C, reserves the product, and records the source of the inquiry. Later, when the manager changes the document status, the change is automatically sent to the client as a message. Thus, messaging and accounting no longer live separately, and employees stop manually duplicating the same data in two places.

An important detail often overlooked: 1C should not be exposed directly to the internet. There should always be an intermediary layer between the messenger and the accounting system that ensures security, limits available operations, and stores keys. This layer also smooths out speed differences: the messenger expects a response in seconds, while a heavy report in 1C may take longer to calculate, so such operations are moved to background processing, and the client is immediately informed that the request is accepted.

LIMITATIONS AND WEAKNESSES

An honest discussion of drawbacks is more important than a list of features, so let's gather the weak points directly.

First — dependence on Meta's rules. Message categories, rates, and moderation conditions change, and the transition to per-operation payment in July 2025 clearly demonstrated this. A budget calculated under the old model could increase, and one should anticipate that the rules will change again.

Second — the channel is no longer the undisputed leader in Kazakhstan. If your audience has shifted to Telegram, investing all efforts only in WhatsApp is short-sighted: it is wiser to build the bot's logic so that it can be reused in multiple messengers.

Third — moderation and blocking. Templates undergo verification, and attempts at mass mailings outside the rules lead to number blocking. Recovery takes time and is sometimes impossible, so gray schemes are a risk for the entire channel, not a way to save.

Fourth — the cost of initiated dialogues. Marketing templates are now charged per message, and with large mailing volumes, the bill grows quickly. Without calculating the economics per message, it is easy to go into the red.

Fifth — the illusion of full automation. A bot without a well-thought-out handover to a live operator quickly irritates clients and spoils the impression more than its absence. The escalation scenario should be designed from the start.

Sixth — integration debt. A bot not connected to accounting becomes another disparate channel where managers manually transfer applications. Without a connection to 1C, automation is half-baked and over time creates more work than it saves.

PRACTICAL CONCLUSION

A specialist setting up a bot should start with the official Cloud API and design scenarios around the free 24-hour window, while heavy requests to 1C should be moved to background processing through an intermediary layer on OData. This provides a predictable price and stable architecture.

A department head should calculate the economics under the new model: how many dialogues the business initiates with paid templates and how many clients start for free. Shifting the load towards client inquiries is the quickest way to reduce costs without losing quality.

An owner should look beyond one channel. WhatsApp is useful, but in Kazakhstan, it is no longer the sole leader, so the bot's logic should be built to transfer to other messengers, and messaging must be linked to the accounting system. A bot detached from 1C is an expense without return.

FREQUENTLY ASKED QUESTIONS

How much does a message to a client cost after July 2025?
It depends on the category and country. Marketing templates are significantly more expensive than service and authentication ones, and responses within the client-opened 24-hour window are free. Therefore, the final amount is determined not so much by the rate as by the structure of your scenarios.

Can I send messages to my database through a regular number?
Through a free app on a phone — no, mass automatic mailings outside the official API violate the rules and lead to blocking. The legal way is Cloud API and agreed templates, where the recipient has a reason to expect the message.

Is it necessary to connect the bot to 1C?
Not technically necessary, but without it, much of the benefit is lost. When the bot sees current stock and statuses from 1C, it responds accurately and without human involvement. Without the connection, employees manually transfer data, and automation remains incomplete.

Will the bot replace operators?
No. It handles typical repetitive questions and unloads the first line, but complex inquiries need to be passed to people. The best results come from a combination where the bot handles routine, and the person handles non-standard situations.